2 ms·
My ex-neighborhood (Magnolia in Seattle) has had Flock Safety install 6 "AI" camera systems, two at each of the three entry/exit point of the neighborhood. Each
by Wistar 7y ago
My ex-neighborhood (Magnolia in Seattle) has had Flock Safety install 6 "AI" camera systems, two at each of the three entry/exit point of the neighborhood. Each installation has a camera overseeing entering traffic and one overseeing exiting traffic. The cameras have license-plate recognition and object detection for car color/make/model, pedestrians, bicyclists and even animals. Flock Safety charges $2,000 per year per camera. Magnolia is one of about 10 Seattle-area neighborhoods with Flock Safety cameras installed. The Magnolia residents themselves pay for the cameras and service. They all can log in to a web service and see the footage and ALPR lists, timestamps, etc.
Friends I have who live there are thrilled with the system although the Seattle police seem to have little interest in the system telling the residents that they are only interested in seeing footage that shows suspicious activities or is directly related to a crime.
https://www.flocksafety.com/ https://www.flocksafety.com/
Porch pirates are so bad in my new neighborhood that I admit to wanting one of these on my dead-end culdesac and have investigated what it would take to build such a system myself.
- knicholes 7y ago1. Set up some raspberry pis with cameras that run a webserver that when hit, it returns the current image. 2. Create a backend that queries each camera for their current image. 3. Run the images through some object recognition neural network (fastai has some great object recognition tutorials). 4. Make your backend stitch together the images from each camera to make a video of what was being recognized when objects of interest were detected (each camera will create its own video) 5. Create a nice little UI/App for everyone to view the movies with some filtering in place (time of day, object(s) of interest). I want to create something like this as well. All of the pieces seem to be pretty straightforward to me except for figuring out power and network connectivity. I guess each neighbor would have to configure the device to use their SSID... and hope someone doesn't steal the raspberry pi and get the SSID credentials from memory... Maybe just make the neighbors use a separate network?
- Wistar 7y agoWow. Thank you. There is an OpenALPR project for license plate recognition. https://github.com/openalpr/openalpr https://github.com/openalpr/openalpr Also at: http://doc.openalpr.com/opensource.html http://doc.openalpr.com/opensource.html And a ready-to-go version you can buy/subscribe based on their stuff: http://www.openalpr.com/ http://www.openalpr.com/